[mapserver-dev] mappostgis.c

[Paul Ramsey]

Performance is next up after I'm sure of correctness. I'll just test
trunk-vs-5.2 and see if I can get trunk running 2x as fast. Then I'll
know I have a winner :)

P

On Tue, Nov 4, 2008 at 10:54 AM, Steve Lime <Steve.Lime at dnr.state.mn.us> wrote:
> A good threat, probably long overdue. Do you have any feel for the performance end of things? I
> recall the Geoserver tests showing GeoServer performing substantially better than MapServer w/regards
> to PostGIS.
>
> Steve
>
>>>> On 11/4/2008 at 11:27 AM, in message
> <30fe546d0811040927n51812e1do272f2cb1bac836b4 at mail.gmail.com>, "Paul Ramsey"
> <pramsey at cleverelephant.ca> wrote:
>> FYI, as threatened, I have updated the postgis driver in trunk rather
>> substantially.  The binary cursor is gone, the SQL generation is a bit
>> more modular, much of the code duplication is removed, some memory
>> leaks are removed, I'm sure some others have been added (valgrinding
>> is still on the to-do list). In general the complexity is a lot lower
>> without the binary cursor, so it's all worthwhile.
>>
>> I put some prototypes and structs in mappostgis.h, so I could bring
>> them into my unit tester more easily. The other thing on my "unit
>> testers wishlist" would be having the mapserver Makefile put things
>> like the INCLUDE paths and LIB flags and -D lists in a place where
>> other Makefiles could import them. For external apps making use of
>> libmapserver.a, it would be quite handy.
>>
>> mappostgis.c has received light testing in map, query, nquery,
>> itemquery and itemnquery modes and seems to hold up OK. (Watching the
>> debug log during the nquery mode is especially horrifying, I had
>> forgotten the full extent of the inefficiency of the current set-up.)
>>
>> If you happen to have PostGIS applications around, I would be
>> interested to hear if things still work "as expected" in trunk.
>>
>> I did add in support for maxfeatures as well, but will probably have
>> to alter it a bit if we do an RFC on paging, since both query and map
>> mode are respecting maxfeatures as it stands.
>>
>> P.
>> _______________________________________________
>> mapserver-dev mailing list
>> mapserver-dev at lists.osgeo.org
>> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>
>